GeForce GT 730 random colorful X crashes

Graphics / Linux Linux
1

Hi there! I have a Desktop PC with a GeForce GT 730.
I tried various different nvidia drivers: 340.76, 340.93, 346.72, 346.82, 352.41, 355.11 and all of them produce the same random crashes.
It does not only crash the X window but the keyboard becomes unresponsive (can’t switch to ttys, can’t strg+alt+del reboot)

Via an ssh session I was able to aquire these dmesg lines from one such crash:

and also really strange is that the broken graphic output is persistent when I use “sudo reboot” which produces this dmesg output after reboot:

help…? is this a bug in the drivers or is this a hardware failure?

I’ve also asked for help with this problem here:
http://forums.linuxmint.com/viewtopic.php?f=59&t=206140

2

My guess would be a hardware problem, given the severity and variety of errors you’re seeing. Please make sure that the graphics card is properly seated and that there is sufficient airflow to keep the GPU cool. Do you have a different GPU you could try, by any chance?

3

Thanks for the reply.

First: I have found that using the open source Nouveau driver included in the Ubuntu 15.10 beta2 release that the system runs completely stable. It gives much lower performance in GPU benchmarks then the original Nvidia driver but nothing noticeable in normal (non 3D gaming) usage.

Second: If it’s a hardware problem, it is not triggered by too high temperatures. The error happens even when the temperature is around 45°C. It also only happens when I use the local displays - When I only use the machine remote, using CUDA I can put heavy load on the GPU for days without crashing.

Third: I’ve only used Windows 10 with the Windows Nvidia drivers on this machine for a few hours but it ran completely stable for those hours. That might have been luck (maybe I’ll try Windows again for a longer period to confirm) but I think it tells me that the problem is only caused by the combination of my Card and the Linux Nvidia drivers.

4

Okey, so I’ve looked at Windows 10 again and found that I get the same crashes when using this software:

to test the graphic cards memory.
So I’m quite convinced now that the gpu is defective and asked the seller to exchange it for another one.